bandit_posterior: bandit posterior

In StatsWithR/statsr: Companion Software for the Coursera Statistics with R Specialization

Description

Utility function for calculating the posterior probability of each machine being "good" in two armed bandit problem. Calculated result is based on observed win loss data, prior belief about which machine is good and the probability of the good and bad machine paying out.

Usage

bandit_posterior(
  data,
  prior = c(m1_good = 0.5, m2_good = 0.5),
  win_probs = c(good = 1/2, bad = 1/3)
)

Arguments

data	data frame containing win loss data
prior	prior vector containing the probabilities of Machine 1 and Machine 2 being good, defaults to 0.5 and 0.5 respectively.
win_probs	vector containing the probabilities of winning on the good and bad machine respectively.

Value

A vector containing the posterior probability of Machine 1 and Machine 2 being the good machine.

See Also

bandit_sim to generate data and plot_bandit_posterior to visualize.

Examples

data = data.frame(machine = c(1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L), 
                  outcome = c("W", "L", "W", "L", "L", "W", "L", "L", "L", "W"))
bandit_posterior(data)
plot_bandit_posterior(data)

StatsWithR/statsr documentation built on Jan. 24, 2021, 10:12 a.m.